Computers and the Humanities

, Volume 10, Issue 5, pp 287–296 | Cite as

The use of the computer in the analysis of German folksongs

  • Wolfram Steinbeck


Computational Linguistic 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    The terms “deutsches Volkslied,” “deutsches volkstümliches Lied,” and “deutsches Lied” have all been subsumed under the rubric “German folksong.” The term “song” or “folksong” in this study refers only to the melody. For simplicity's sake, the term “German folksong” is used here in a broad sense to include all printed songs with German texts which have been or still are in popular use. To this definition belong the type of folksong found in the collection of Erk-Böhme (Deutscher Liederhort, ed. by L. Erk, Berlin 1856, rev. and expanded by Fr. M. Böhme, 3 vols., Leipzig, 1893–4) as well as the popular art songs included in the collection of Böhme (Volksthümliche Lieder der Deutschen im 18. und 19. Jahrhundert, ed. Fr. M. Böhme, Leipzig, 1895).Google Scholar
  2. 2.
    See, for example, B. J. Riepel,Anfangsgründe zur musicalischen Setzkunst… Erstes Capitel De Rhythmopoeia, Oder von der Tactordung. Regensburg and Vienna, 1752, p. 23f.; H. Chr. Koch,Versuch einer Anleitung zur Composition, 3 vols., Vol. I: Rudolstadt, 1782, Vols. II–III: Leipzig, 1787–93; Vol. II, p. 366, and Vol. III, p. 53.Google Scholar
  3. 3.
    Folksong researchers and catalogers in eastern European countries, for example, have long made use of the computer. See, e.g., B. K. Csébfalvy et. al., “Systematization of Tunes by Computers,”Studia musicologica 7 (1965); A. Elscheková, “General Considerations on the Classification of Folk Tunes,”ibid. Google Scholar
  4. 4.
    See note 1. These restrictions are based on the need for an homogenous base of material which avoids the philological, genetic, and folkloristic problems common to folksong research. Printed melodies, by their very nature unambiguous, present us with none of these problems.Google Scholar
  5. 5.
    See, for example, W. Wiora,Die deutsche Volksliedweise und der Osten, Wolfenbüttel and Berlin, 1940, esp. pp. 9ff.; idem,Europäische Volksmusik und abendländische Tonkunst, Kassel, 1957 =Die Musik im alten und neuen Europa. Eine Schriftenreihe hrsg. von W. Wiora, Vol. I, esp. p. 117ff.Google Scholar
  6. 6.
    See W. Suppan,Volkslied. Seine Sammlung und Erforschung, Stuttgart, 1966, pp. 17ff.Google Scholar
  7. 7.
    See, for example, St. Bauer-Mengelberg, “The Ford-Columbia Input Language,” inMusicology and the Computer: Three Symposia, ed. by Barry S. Brook, New York, 1970; Barry S. Brook, “The Simplified ‘Plaine and Easie Code System’ for Notating Music…,”Fontes Artis Musicae, 12, 1965; Murray J. Gould and George W. Longemann, “ALMA: Alphanumeric Language for Music Analysis,” inMusicology and the Computer, op. cit. Jerome Wenker, “MUSTRAN II: A Foundation for Computational Musicology,”Computers in the Humanities, ed. by J. L. Mitchell, Edinburgh, 1974.Google Scholar
  8. 8.
    There would be no need in this study, for example, to code any of the divergent phrasing, tempo, or dynamic markings often encountered in different editions of the same folksong.Google Scholar
  9. 9.
    See W. Steinbeck, “ANTOC. Ein neues Verfahren der Aufzeichnung von Melodien durch den Computer,”Report of the International Musicological Congress, Berlin 1974 (in press).Google Scholar
  10. 10.
    See below p. 291.Google Scholar
  11. 11.
    E.g., it is necessary to record the tonic note and its octave range because of the transposition of all melodies to a single key for purposes of comparison.Google Scholar
  12. 12.
    If the machine finds neither D nor M (symbols for major [Dur] and minor [Moll] respectively), but, for example, another prearranged letter instead, a special program function is automatically called upon. In this way, ANTOC is capable of handling those melodies which cannot be classified as either major or minor, or those melodies which do not adhere to bar-line principles.Google Scholar
  13. 13.
    Normally, data input is accomplished through magnetic tape or punch cards. These processes entail both a high risk of error because of the multiple stages of transcription and time-consuming correction procedures. A particularly fast and efficient method, described by K. Csébfalvy et al (op. cit.), uses a specially made instrument, the Klaviatur, which allows for the direct input of melodies with the option of simultaneous acoustical control.Google Scholar
  14. 14.
    D. F. Scheurleer, “Welches ist die beste Methode, um Volks- und volksmäßige Lieder nach ihrer melodischen … Beschaffenheit lexikalisch zu ordnen?,”Festschrift der Internationalen Musikgesellschaft, 1(1899–1900).Google Scholar
  15. 15.
    H. Mersmann, “Grundlagen einer musikalischen Volksliedforschung,”Archiv für Musikwissenschaft 4 (1922), p 145ff.Google Scholar
  16. 16.
    G. Birkner, “Eine musikalische Katalogisierung des neueren deutschen Volksliedes,”Zeitschrift für Volkskunde 60 (1964); J. Lansky and W. Suppan, “Der neue Melodien-Katalog des Deutschen Volksliedarchivs,”Fontes Artis Musicae 10 (1963).Google Scholar
  17. 17.
    The system of the Deutsches Volksliedarchiv, for example, includes under each entry the source of the song, formal scheme, locale, and related literature. The system of cataloging by accented notes falls short for purposes of determining similarity between melodies, however. Only one of the accented notes need be different for basically similar melodies to be classified far apart from one another.Google Scholar
  18. 18.
    See, inter alia, L. A. Hiller et al., “Automated Music Printing,”Journal of Music Theory 9 (1965).Google Scholar
  19. 19.
    See, for example, N. Böker-Heil, “Weiße Mensural-notation als Computer-Input und -Output,”Acta Musicologica 43 (1971); idem., “Plotting Conventional Music Notation,”Journal of Music Theory 16 (1972); D. Byrd, “A System for Music Printing by Computer,”CHum 8 (1973/74); H. B. Lincoln, “The Current State of Music Research and the Computer,”CHum 5 (1970/71).Google Scholar
  20. 20.
    On the line printer in Kiel, for example, which produces 1000 lines per minute, at least 30 notes with all accompanying symbols can be produced in just under one second. For an example of LPTN, see Fig. 2 at the end of this article. See also H. Küpper, “Computer Komponieren,”Umschau 22 (1967).Google Scholar
  21. 21.
    See esp. W. Wiora, “Elementare Melodietypen als Abschnitte mittelalterlicher Liedweisen,” Miscelánea en homenaje a Monsenor H. Anglés, Vol. II, Barcelona, 1958–61.Google Scholar
  22. 22.
    See esp. Ibid., “Elementare Melodietypen als Abschnitte mittelalterlicher Liedweisen,” Miscelánea en homenaje a Monsenor H. Anglés, Vol. II, Barcelona, 1958–61, p. 994.Google Scholar
  23. 23.
    See W. Wiora, “Zur Methode der vergleichenden Melodienforschung,”Methoden der Klassification von Volksliedweisen, Bratislava, 1968, p. 38.Google Scholar
  24. 24.
    Op. cit., p. 41: “In the building of a house, the same bricks go into the construction of different parts of the house and are eventually hidden from view by the plaster. In a similar manner, the ‘bricks’ or ‘building blocks’ of a folksong may be arranged in any number of ways and eventually be ‘covered’ by text and melody.”Google Scholar
  25. 25.
    Each time-value is multiplied to accommodate the smallest time-value encountered in the material, the sixteenth-note triplet. A quarter note, for example, is six times the duration of this smallest value, a sixteenth-note, one-and-a-half times the value. These durations are then calculated as a fraction of a measure, according to the meter. One note of the sixteenth triplet would thus make up 1/12 of a measure in 2/4 time, or 1/9 of a 3/8 measure. The lowest common denominator becomes 144, in which the sum of all note-values within a measure is 144/144=1. The values of individual notes within a measure can thus be easily calculated and compared: in a 2/4 measure, the sixteenth-triplet note takes on the value of 12/144; a quarter note in a 4/4 measure becomes 36/144, etc.Google Scholar
  26. 26.
    To simplify the programmed analysis of the melodies, all material, once it has been transposed, is automatically stored in a purely numerical code. This code indicates absolute pitch by a number sequence beginning with 1.Google Scholar
  27. 27.
    See below p. 291. The comparison of identical attributes in different melodies without regard to their relative significance is bound to produce misleading results. Such a method has been put forward by R. Kluge (“Faktorenanalytische Typenbestimmung an Volksliedmelodien” = Beitrag zur musikwissenschaftlichen Forschung in der DDR), (1974). Kluge himself recognizes the shortcomings of his system in an admirable piece of self-criticism at the end of his study: “Certain attributes seem to differ in factors of importance. … These differences are necessarily reflected in the determination of similarity” (p. 156). See also the extensive catalog of attributes (op. cit., p. 1ff), which Kluge, again self-critically, questions on the bases of “completeness and “appropriateness.” For further remarks on Kluge's research, see also Note 30.Google Scholar
  28. 28.
    I am most thankful to Prof. Dr. Fr. Vogel of the Institut für Statistik und Ökonometrie at the Universität Kiel for his valuable assistance in the area of cluster analysis. See also P. E. Green and V. R. Rao, “A Note on Proximity Measures and Cluster Analysis,”Journal of Marketing Research 6 (1969); N. Jardine and R. Sibson,Mathematical Taxonomy, London (1971); D. G. Morrison, “Measurement Problems in Cluster Analysis,”Management Science 13 (1967); Fr. Vogel,Probleme und Verfahren der numerischen Klassifikation…, Göttingen (1975).Google Scholar
  29. 29.
    The problem of applying cluster analysis to different classes of attributes (see above, p. 289) is methodologically problematical. The attributes should therefore belong to one class. Fortunately, this restriction presents no problem for the material under consideration here. Alternative attributes may be eliminated from the similarlity determination process, as their characteristics can be expressed through discontinous attributes. For example, mode (an alternative) can be determined through the sequence of notes (a discontinuous attribute).Google Scholar
  30. 31.
    H. Riemann,Grundriß der Kompositionslehre (Musikalische Formenlehre) 5th ed., Berlin and Leipzig, 1916/17.Google Scholar
  31. 32.
    The image of the building blocks also illustrates the inherent formality of these melody segments. Many segments appear again and again in numerous melodies in varying rhythmic contexts. The appropriateness of this building block-image is further reflected in the “dice music” (“Würfelmusik”) popular in the second half of the eighteenth century (see, for example, J. Ph. Kirnberger,Der allezeit fertige Polonaisen- und Menuettenkomponist, Berlin, 1757). A detailed analysis through the aid of the computer of this type of music will be published at a later data.Google Scholar
  32. 33.
    See W. Steinbeck, op. cit., p. 6.Google Scholar
  33. 34.
    The results from this method are admittedly not as refined as those from cluster analysis. Because of its simplicity, though, VETTA is easier and more rapid to use in determining similarity. Rapid retrieval of information is a factor of no small significance in a melody catalog.Google Scholar
  34. 35.
    See, for example, W. Wiora,Europäischer Volksgesang: Gemeinsame Formen in charakteristischen Abwandlungen = Das Musikwerk, vol. 4, Cologne, 1952.Google Scholar
  35. 36.
    VETTA is thus applicable only to melodies transcribed to indicate bar-lines and measures, as is the present body of folksongs. Were another suitable criterion for juxta-position established, an appropriate method could be developed for melodies not transcribed in this manner.Google Scholar
  36. 37.
    A 4/4 measure, for example, has the size of four quarter-notes (=1), a 3/4 measure the size of three-quarter notes (=0.75).Google Scholar
  37. 38.
    The number of beats per measure need not necessarily be identical with the numerator of the time-signature. This is especially true for folksong melodies, which are transcribed from oral traditions. See C. Dahlhaus, “Was ist musikalischer Rhythmus,” inProbleme des musiktheoretischen Unterrichts, Berlin (1967), = Veröffentlichung des Instituts für Neue Musik und Musikerziehung Darmstadt, Vol. VII.Google Scholar
  38. 39.
    The VETTA Method may be changed by decreasing the size of the comparison fields to correspond with the size of the smallest transposed time-value. Longer time-values would then be comparable through a combination of comparison fields (see Fig. 1, Melody 1, m. 2, where a note encompasses two beats, indicated by the joining of the two blocks). By thus including every note, VETTA can identify identical melody segments unequivocally. This method, while feasible, would be extremely time-consuming.Google Scholar
  39. 40.
    H. Mersmann, op. cit.,Archiv für Musikwissenschaft 5 (1923), 132f.Google Scholar
  40. 41.
    Cf. the cataloging principle of the melody catalog in the Deutsches Volksliedarchiv in Freiburg (above, p. 288).Google Scholar
  41. 42.
    See above, Note 1.Google Scholar

Copyright information

© Pergamon Press 1976

Authors and Affiliations

  • Wolfram Steinbeck
    • 1
  1. 1.Institute of MusicologyUniversity of KielGermany

Personalised recommendations